Enterprise AI Costs and Governance Put Scaling Ambitions Under Pressure Across APAC

Cloudera’s 2026 report finds that 89% of enterprises have limited or redesigned AI use cases because of rising token consumption, while businesses across Asia Pacific struggle to translate increased investment into measurable returns.

Enterprise investment in artificial intelligence is continuing to grow, but the economics of deploying AI at scale are becoming harder to justify. Across Asia Pacific (APAC), 58% of enterprises increased their AI investments over the past year, yet only 28% said their AI initiatives currently generate more value than they cost. Another 21% reported that AI costs outweigh the value delivered, highlighting the growing gap between adoption and commercial returns.

The challenge extends beyond initial implementation. According to Cloudera’s State of Enterprise AI 2026: The New Rules of Scale, released on October 27, 89% of global respondents have limited or redesigned AI use cases because of rising token consumption. The findings suggest that enterprises are reassessing not only how extensively they deploy AI, but also the infrastructure, governance and operating costs required to sustain those deployments.

The report, commissioned by Cloudera and conducted by Wakefield Research, surveyed 1,500 IT leaders and decision-makers globally in September 2026. Its findings point to a shift in enterprise AI priorities, from expanding deployments towards controlling costs, managing operational risks and demonstrating measurable business value.

Rising inference costs are changing deployment strategies

As enterprises move AI applications from experimentation into production, the volume and complexity of model interactions can increase substantially. Token consumption, particularly in applications involving long context windows, repeated model calls and multi-step workflows, can create variable operating expenses that are difficult to predict.

Cloudera’s survey found that 45% of global respondents had moved AI workloads in response to higher or more volatile costs. Across APAC, 26% had gone further by pausing or canceling AI use cases because of cost volatility. These decisions indicate that cost considerations are beginning to influence application architecture and deployment choices. Enterprises may need to reconsider model selection, processing frequency, infrastructure environments and the amount of data supplied to AI systems.

The challenge is not simply reducing the cost of individual model calls. Businesses must assess whether the overall cost of an AI application is justified by the productivity improvements, operational efficiencies or revenue opportunities it generates.

Governance and reliability complicate enterprise adoption

Financial considerations are only one part of the scaling challenge. The report found that 75% of global respondents identified governance and compliance as constraints on scaling AI, compared with 73% in APAC. Furthermore, 55% globally and 56% across APAC reported that governance and compliance costs had increased faster than expected.

Reliability also remains a concern as AI systems enter business-critical environments. According to the survey, 98% of respondents had pulled back at least one AI deployment because of failures, hallucinations or unexpected behavior. These findings highlight the difference between demonstrating that an AI application works under controlled conditions and maintaining its reliability across real-world business operations. Enterprises must account for data access, model behavior, security, auditability and human oversight throughout deployment.

Singapore’s regulatory approach reflects this growing emphasis on operational accountability. Its Infocomm Media Development Authority updated the Model AI Governance Framework for Agentic AI in May 2026, adding practical guidance on multi-agent systems, third-party risks and automation bias. The framework emphasizes human accountability alongside technical safeguards as organizations deploy increasingly autonomous AI applications.

The location of AI processing is also becoming a strategic infrastructure decision. Cloudera found that 84% of global respondents and 79% across APAC said data residency, sovereignty or jurisdiction-specific requirements influence where their AI workloads operate. Additionally, 58% globally said relocating workloads requires moderate or major redesign.

For enterprises operating across multiple markets, infrastructure decisions must account for different regulatory requirements, data sensitivity, performance needs and operating costs. This is particularly relevant in APAC, where organizations frequently manage data across jurisdictions with different legal and compliance frameworks.

The report identifies resilience as the leading reason for operating AI workloads across multiple environments, cited by 42% of respondents. Sovereignty and residency requirements and cost reduction were each cited by 38%, while 37% pointed to proximity to data. This creates demand for infrastructure that allows organizations to choose where workloads run without repeatedly rebuilding their applications and governance systems. Hybrid environments, combining public cloud, private infrastructure and on-premises deployments, can provide greater flexibility, although managing consistent controls across these environments remains a challenge.

Building the foundations for sustainable AI

The report also highlights areas enterprises believe they should have prioritized earlier. If given the opportunity to restart their AI journeys, 49% of respondents said they would strengthen governance from the beginning, 46% would invest in data quality earlier, and 45% would modernize infrastructure sooner. Another 45% would allocate more resources to employee training. Skills are a particular concern across APAC, where 47% of respondents identified a lack of expertise as a constraint on scaling AI.

These findings suggest that sustainable adoption depends on more than access to models and computing capacity. Data quality, employee capabilities, infrastructure readiness and governance must develop alongside AI deployment. Organizations that postpone these investments may encounter higher costs and greater operational complexity as applications expand.

Remus Lim, Senior Vice President, Asia Pacific and Japan at Cloudera, said enterprises need to focus on extracting greater value from their existing data and infrastructure.

“To relieve these pressures, organizations need to move beyond simply expanding AI deployments to focus on maximizing their existing data and infrastructure. By bringing AI directly to data wherever it resides, enterprises can minimize unnecessary data movement, optimize workload placement based on business needs, and maintain consistent governance across environments,” Lim said.

For enterprises across APAC, the next stage of AI adoption will depend on whether investment can translate into sustainable business outcomes. As costs, regulation and infrastructure requirements evolve, deployment volume alone will be an inadequate measure of progress. The more significant test will be building AI systems that remain reliable, governable and economically viable as their use expands.

Quick Takeaways

• 89% of global enterprises have limited or redesigned AI use cases due to rising token consumption.
• 58% of APAC enterprises increased AI investment, but only 28% say AI currently generates more value than it costs.
• 26% of APAC respondents have paused or canceled AI use cases because of cost volatility.
• Governance and compliance remain major scaling constraints, cited by 73% of APAC respondents.
• 79% of APAC enterprises say data residency, sovereignty or jurisdiction-specific requirements influence AI workload deployment.
• 47% of APAC respondents identify skills and expertise as barriers to scaling AI.
• Enterprises are prioritizing better data quality, stronger governance and infrastructure modernization to make AI deployments more sustainable.